Detecting and extracting method and system characterized by image dimension not transforming

A technology of scale-invariant features and extraction methods, applied in character and pattern recognition, instruments, computer parts, etc., can solve the problems of poor robustness, low accuracy and stability, and achieve high accuracy and stability, good robustness

Inactive Publication Date: 2014-05-28
SUZHOU UNIV
View PDF2 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Benefits of technology

This technology allows us to define an orientation matrix called LT without changing its characteristics over time when converting images from one resolution level (pixels) into another. By calculating this vector's values at different angles around every single frame, we can accurately determine how well these orientations are used during conversion between two levels or even across multiple frames.

Problems solved by technology

The technical problem addressed in this patented text relates to improving the precision and reliability of finding interesting areas within digital imagery (DIG) scenes that may be affected by different factors like scaling transforming from one size to another shape over time).

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Detecting and extracting method and system characterized by image dimension not transforming
  • Detecting and extracting method and system characterized by image dimension not transforming
  • Detecting and extracting method and system characterized by image dimension not transforming

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0067] Embodiment 1 of the present invention discloses a method for detecting and extracting image scale-invariant features, please refer to figure 1 , the method includes:

[0068] S1: Perform mathematical model transformation on the two-dimensional image to obtain a mathematical model of each pixel in the two-dimensional image.

[0069] Due to the analysis of the image, the gray value of the image is generally not processed directly, but the image is processed by some mathematical or geometric transformation methods, such as Fourier transform, small filter transform and Gabor filter. Based on this, the present invention assumes that the polynomial coefficients can sufficiently express the local characteristic information of the two-dimensional image, and transforms the image based on the multinomial expansion formula to obtain the mathematical model of the local structural characteristics of the image. Specifically, each pixel of the image is expanded in the form of polynom...

Embodiment 2

[0161] Embodiment 2 of the present invention discloses a detection and extraction system for image scale-invariant features, please refer to Figure 9 , the system includes a mathematical model transformation module 100 , a feature scale acquisition module 200 , a direction tensor acquisition module 300 , a judgment module 400 and an interest point acquisition module 500 .

[0162] The mathematical model conversion module 100 is configured to perform mathematical model conversion on the two-dimensional image to obtain a mathematical model of each pixel in the two-dimensional image.

[0163] The characteristic scale obtaining module 200 is configured to obtain the characteristic scale of each pixel from a preset scale space, the scale space includes N different scales, and N is a natural number greater than 1.

[0164] Wherein, the feature scale acquisition module 200 includes a calculation unit and a feature scale acquisition unit. The calculation unit is used to calculate th...

Embodiment 3

[0189] In the third embodiment, the method of the present invention is evaluated by extracting the points of interest of the actual scene image. Specifically, the repetition rate criterion is used to evaluate the method of the present invention.

[0190] The repetition rate criterion is mainly to evaluate whether the location and scale of the detected and extracted interest points are accurate. For a pair of images, the repetition rate mainly refers to the ratio of the number of repeated interest points determined by the homography matrix in two images with a certain transformation to the total number of interest points in an image with fewer interest points. The two images correspond to the point of interest X a and x b The matching conditions are as follows:

[0191] a. The positioning error of the two images corresponding to the point of interest is less than 1.5 pixel values, namely: ||Xa-HX b ||<1.5.

[0192] Among them, H represents the homography matrix.

[0193] ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a detecting and extracting method and system characterized by an image dimension not transforming. According to the method, a direction tensor T not transforming along with transformation of the dimension is defined for each pixel point according to a mathematic model and a characteristic dimension of each pixel point in a two-dimension image, the smallest characteristic value of the direction tensor is calculated, and the characteristic energy intensity of the pixel point is presented through the smallest characteristic value in the same direction; finally, whether the pixel point is an interesting point is detected by judging whether the smallest characteristic value of the pixel point is the largest value of a preset window, wherein the largest value comprises the smallest characteristic value of the pixel point, and extraction of the interesting point is achieved. According to the detecting and extracting method and system characterized by the image dimension not transforming, due to the fact that the direction tensor T not transforming along with transformation of the dimension is defined for each pixel point, the smallest characteristic value of the direction tensor T does not transform along with transformation of the dimension, wherein the smallest characteristic value presents the characteristic energy intensity of the pixel point. Therefore, on the basis of an angle of the characteristic energy intensity finally, the interesting point extracted through the smallest characteristic value of the direction tensor T has good robustness, high accuracy and stability for transformation of image dimension.

Description

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Owner SUZHOU UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap